An assistant must use rapidly changing private policies. The team proposes fine-tuning as the only update mechanism. Which alternative should be evaluated first for factual freshness?

1. Use a larger output limit as a knowledge update
2. Retrieval of current authorized source material
3. Assume one fine-tune learns all future policies automatically
4. Remove citations to avoid visible staleness